Dave Snowden

Typology or Taxonomy?

RSS Feed

EFE28134-0F76-4200-B82F-625BE8D1020E.jpg The Academy of Management award for my HBR cover article with Mary Boone says among other things that the paper introduces complexity science (please note the use of “science” which will be important later) and further that it applies this perspective to advance a typology of contexts to help leaders to sort out the wide variety of situations in which they must lead decisions. Now the key word here is typology and I used that myself as the title of yesterday’s post. Those who have heard me speak, or have listened to more than one or two of my podcasts will have heard me suggest that the word taxonomy not only rhymes with taxidermy but in the hands of many a KM practitioner produces the same effect.

Now in some ways this is a bit unfair, there is a clear role for taxonomies in many fields including knowledge management. The problem is when they taken to excess or used inappropriately. A search on “bad taxidermy” produces some very scary pictures, one of which I selected for this blog, but its nothing to what I have seen in many a KM programme. The issue is the general issue with categorisation is a situation where there is substantial change. I remember in IBM the taxonomy in the KM system took three years to catch up with the use of story telling, and by the time it did there was a clear distinction between story telling and micro-narrative. It still hadn’t caught up with that change by the time I left by the way and I suspect it never will.   

There is no clear agreement on the definitions of typology and taxidermy, well there are in biology but as we move into other fields it gets harder. I’d also make it clear that I am pushing the boundaries of my knowledge here so if Patrick Lambe (or someone else who does understand this) wants to correct me on my use that is fine. To be honest I am less concerned about the labels than I am about the distinctions. Knowledge management is closer to policy formation than it is to biology so I am going to use the definitions in Kevin Smith’s September 2002 paper in the Policy Studies Journal Typologies, taxonomies, and the benefits of policy classification as my starting point.

  • In a typology the dimensions represent concepts, they do not necessarily exist in physical reality (although they can). As such typologies generate heuristics which are more adaptive under changing circumstances. On the downside the concepts can be arbitrary may not be exhaustive and can easily be subject to clashes of interpretation.
  • A taxonomy on the other hand classifies things based on clear empirical characteristics and will have rules that allow determination of location. They have clear boundaries mostly determined by cluster analysis allow rapid decision making . On the downside, once a taxonomy is established if something does not fit, it will be made to fit as the taxonomy itself creates a filtering mechanism through which we filter observable characteristics.

Now in policy and in knowledge management, as applied to decision making, distinctions are the main socially constructed, subject to large variation in differing contexts and are difficult to subject to empirical validation. Names of course play a major role in human sense-making and carry with then historical meaning and implications for action. Names carry with them power, as any reader of myths and legends will tell you. The problem with categories is the power of the name can result in perverted response (think of the different uses of “Jew” over the centuries) or can result in novelty (both threat and opportunity) being missed as it doesn’t fit.

I experienced that problem in IBM. We were developing a whole new concept of research in services, modeled on medical science, in which concepts and practice co-evolved. That meant you sold consultancy around intractable problems, then applied sound theory in a series of safe-to-fail experiments, modify practice as you went along. This created a major issue for IBM. We couldn’t sit in research as we were also consultants; we couldn’t belong to consultancy as we didn’t focus on one offering in one industry sector in one country with minimum gross margin with formal bid processes. Net result we ended up in Marketing as they found us interested, and also amusing when we won a DARPA research contract from the Watson Labs. The irony was even greater as one of the reasons IBM had bought DataSciences (my original company) was precisely because we understood services both in terms of delivery and creation (I must tell the story of the Genus Programme one of these days). They knew their models were manufacturing ones, but they couldn’t abandon them until the failures were multiple and progressive and even then they found it difficult.

The message is very simple rigid boundaries have huge value in static situations so taxonomies work. But where things are subject to rapid change and the possibility of encountering novelty is high they are plain dangerous. However we do need constructs to make sense of the world and that is were conceptual frameworks, or typologies come into their own.

A Cynefin postscript

Now Cynefin is a bit of a hybrid. It is a conceptual framework so at that level its a typology, but the dimensions are based on natural science so there is an empirical aspect but its not from any form of cluster analysis. I’ve always said that the approach I and colleagues have developed over the years is a form of naturalising sense-making; the naturalising is a philosophical reference to rooting theory in the natural sciences. Now this is useful but it can lead to some confusions if people seize on it and make it a two by two matrix. It can, and is used as such but properly used it is a lot more. Some quick points here and I plan some more posts on aspects of this:

  • Cynefin has five domains; simple, complicated, complex, chaotic and disordered. The disordered domain recognizes the essential inauthenticity to any typology of human experience so disorder is the state of not knowing which of the domains you are in (and you may be in several). Disorder is frequently left out which is a pity.
  • The chaos domain is always a transitionary state and Cynefin is as much about movements between domains (dynamics) as it is about the domains themselves. Now anyone with the right science background gets that quickly, order comes for free to quote John Holland (I think but it might be one of the other Santa Fe founders and I have limited internet access so can’t check that).
  • Its very important to use the constrain based definition of the domain as that allows you to understand how to move between them. If there are no constrains its chaotic. If the constraints are severe enough to make agent behaviour predictable then its ordered, if the nature of those constrains are self-evident then its Simple, if they require analysis then they are Complicated. If the constrains and agents co-evolve then its Complex. That means that changing the constraints can change the nature of the system and consequently the nature of situational assessment and decision making.
  • Simple is next to Chaotic as complacency, or inappropriate application of constrains which can lead to catastrophic failure (again movement)
  • Ideally the framework is social constructed (the boundaries emerge from the data), ideally with the four points method I described in my paper on the history of Cynefin. However it can also be used as a categorisation framework (boundaries precede the data) and the HBR article largely focuses on this aspect as it was an approach more familiar to the readers.

Now The “New Dynamics of Strategy” paper and also “Multi-Ontology Sense-Making” both available from the web site follow this approach in full, the HBR article largely focuses on classification as that aspect was appropriate to the target audience.

So Cynefin is a hybrid. It is a typology in that its dimensions are conceptual, heuristic based and arise from people’s own interpretation of their own situation. But its not a pure hybrid as the dimensions themselves has an empirical base in the natural sciences. It can also be used as a taxonomy, but with care. Again you see the medical model of research, in which soundly rooted science is experimented with in practice both to modify the theory and co-evolve sound practice.

A post post script on the use of Cynefin by others

Now as Cynefin has gained in popularity, recently getting an award as one of the 50 most cited papers in 2007, it has been subject to a lot of use, modification and some misuse. Shawn Callahan’s folksy video on Cynefin which retains its popularity has considerable utility as an introduction. It is not wrong but neither is it fully right as it ignores the disordered domain and to some extend the importance of dynamics. All well and good, Joseph Pelrine has done some excellent work and published papers using the basic Cynefin domains and there are many other examples of useful developments and alternative perspectives; too numerous to mention.

At the other extreme (and about the only occasion on which I have ever intervened) I have seen the name and shape taken and redefined while retaining the name. I’ve never understood people who do this sort of thing. I have acknowledged my debt to Max Boisot’s I-Space but I would never use his representations and names as to do so would be a form of plagiarism which I detest. Now this has only happened with one individual and I now confine myself to patiently making the point from time to time that people should create their own names for their own models rather than trying to catch a lift on a more powerful brand (calling it Cynefin-like really is pathetic by the way.) I’ve moved from mild irritation to a degree of pity to be honest. It seems that anything with four categories from the OODA loop (different versions of OODA would apply in each domain) to Causal Layered analysis (clearly a complicated domain tool) is matched category for category and while such might be explained away as a mash up in the main I think its more of a hashup. I don’t own the words Simple, Complicated, Complex and Chaotic. Variations are used by many others sometimes with acknowledgement, sometimes without, sometimes as a parallel invention. All part of the rich tapestry of ideas. However I do think I have some authority over the use of those terms within the basic form of the Cynefin representation and even more so when the Cynefin name is used.